LV chamber extraction from 3-D CT images--accuracy and precision.

Measurement of left ventricular (LV) chamber volume and shape from three-dimensional (3-D) CT images, generated by the fast X-ray CT scanner known as the dynamic spatial reconstructor, has previously been done using manual slice-image editing. To reduce the exorbitant operator analysis time and operator-dependent measurement variations of manual slice-image editing, we have devised a semiautomatic method for LV-chamber extraction. The method draws upon a minimum requirement for selective manual slice-image editing and mostly makes use of automatic image-analysis operations. Detailed validation results over a wide range of hemodynamic and image-analysis conditions show that the measurements of the semiautomatic method strongly correlate with those made via manual slice-image editing and exhibit a lower intertrial variability. Further, the method reduces operator interaction time by nearly an order of magnitude over that of manual slice-image editing, but provides more detailed 3-D structural definition.